Lstopo doesn't have system memory mappings at low addresses. Make
sure we leave enough GPUVM address space for kernel allocations
(currently only CWSR) before the start of the user-managed SVM
aperture.
Change-Id: Ic197f7bd5a3cfb150a0da2bfdbc848664e7869be
[ROCm/ROCR-Runtime commit: cac0c08496]
Non-canonical GPUVM aperture doesn't exist on dGPUs. Remove comments
and code that say otherwise.
Fix alignment of GPUVM aperture for gfx801. Requires the same workaround
as gfx802. It's not used for anything on gfx801 yet, but will be soon.
Change-Id: I88607fe7b340081cc0715b85f28fdbf5f1bb0ad7
[ROCm/ROCR-Runtime commit: b837c3e7b0]
Use the object size when freeing address space, instead of the
parameter passed in by the caller. The parameter may be incorrect
due to app or runtime bugs, or when the buffers is an AQL ring
buffer with double mapping workaround.
Change-Id: I00bb31d4520ef969a49d6d5ea723e8a33418acc3
[ROCm/ROCR-Runtime commit: 006f3ee41b]
The alignment performed in vm_find_object_by_address isn't sufficient
because it doesn't take into account the offset from the start of the
page.
This fixes a bug where certain unaligned userpointers and sizes fail
to register correctly.
Change-Id: I17872e264467a619f5e1bedb7e1ed3d994a856bf
[ROCm/ROCR-Runtime commit: 8a0161d6bb]
Use the aligned size of the buffer objects for CPU unmapping in
__fmm_release instead of relying on the unaligned size passed in by
the caller.
Change-Id: If986ec24e9a05d32981549fddbf143221fc40bac
[ROCm/ROCR-Runtime commit: 7a383f9d88]
Allocate SVM address space for the registered memory and use new
userptr support in KFD to create a system memory BO associated with
the given user pointer. Map this BO at the SVM address for CPU
access.
MapMemoryToGPU can be used with the registered user pointer and
will return the SVM address as alternate GPUVA.
Change-Id: I4886e193c51fb6870a567878870c36bf8b5c3748
[ROCm/ROCR-Runtime commit: 85f9efb1a0]
Allocations from GPU nodes will return VRAM, not system memory.
Only non-paged allocation from GPU nodes is supported. System
memory can only be allocated from CPU nodes (usually node 0).
The HostAccess flag is no longer used to distinguish the memory
type. It only indicates, whether the memory is mapped for CPU
access.
Maintain compatibility with broken KfdTests by returning system
memory for paged-memory requested from GPU nodes.
Change-Id: I514defede735f55e6de436f41944125b6f2c4ccf
[ROCm/ROCR-Runtime commit: 887b32fe86]
All tonga page size alignment is done in the memory management
functions in fmm.c. All other code only specifies the minimum
alignment it needs and lets fmm.c handle the HW-specific
alignment.
Clean up aligned-exec memory allocation in queue.c to remove
hard-coded TONGA_PAGE_SIZE alignments and remove code duplication.
Make sure alignments are consistent between allocate and free.
Change-Id: Ia8923448173d1cef315af24cebff12adef385cb0
[ROCm/ROCR-Runtime commit: cc9fc386bd]
Stop using NUM_OF_SUPPORTED_GPUS. For now the definitions itself cannot
be removed as ioctl code is in upstream Kernel.
Change-Id: If846625a8ad5062d5483e762850c793d3c00b9d0
[ROCm/ROCR-Runtime commit: ce83dc623f]
Fix hsaKmtRegisterMemory to be a no-op for now and move the multi-GPU
implementation to hsaKmtRegisterMemoryToNodes. Make GPU memory mappings
of host memory visible to all GPUs by default. Device memory is still
visible to the allocating GPU only by default (but can be overridden
with hsaKmtRegisterMemoryToNodes for experimenting with P2P).
Change-Id: I73408afbe3b10c8dad2ab3a780f58413249692e6
[ROCm/ROCR-Runtime commit: 063ad3ad9e]
This is required when we have a debug session
Change-Id: If9d6d2d23a9016b6ca9562e02a91fc16e0354ee4
Signed-off-by: Yair Shachar <Yair.Shachar@amd.com>
[ROCm/ROCR-Runtime commit: 681f4dcecc]
When the Thunk is initialized multiple times in the lifetime of a single process
, some global resources are leaked. This can happen when dlopen and dlclose are
used to load the library at runtime, rather than linking the runtime against
the Thunk. This patch adds the destructor to release global resources when
dlclose is called.
Change-Id: Ia00da0d41f095d0b2706f98c0e75effedd596f49
[ROCm/ROCR-Runtime commit: 582b70f9c3]
This will also fix out of bound access in functions
fmm_get_aperture_base_and_limit and fmm_release
Change-Id: Icf064c46647e69a069126171dbacdf3d5b27f972
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: a4cf02d797]
dgpu_aperture and dgpu_alt_aperture will be shared by all dGPUs.
Change-Id: I814495e43b51acabdc6266cfa8d83db5a062e20d
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 2903a610e1]
Break from the for-loop once dgpu VM range is found, otherwise the
length is reduced by half
Change-Id: Ie602054c16ea69ea1cbb75e804ead551bc3615c0
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 5a55383baf]
Access to reserved address space that has not been allocated should
result in a segfault. Use PROT_NONE to ensure that.
Change-Id: Ic5da9392fabbe78c9ec14f98e8b7b47e5267a98a
[ROCm/ROCR-Runtime commit: 62337b6c0a]
Passing in the wrong aperture resulted in failure to unmap scratch.
Change-Id: Icd7423abfb1bcc773b33becffcbefc233f4ff340
[ROCm/ROCR-Runtime commit: bd93eecc64]
Unify fmm_get_aperture_base and fmm_get_aperture_limit into one
function. Make the return value to HSAKMT_STATUS.
Change-Id: I0b3f563ffb268947ab891f4935f61788d0af0e01
[ROCm/ROCR-Runtime commit: cb53548c89]
hsaKmtAllocMemory only allocates aligned address space and sets up
the scratch_physical aperture to match the allocated address space.
Actual allocation of backing memory happens in hsaKmtMapMemoryToGPU.
Change-Id: Ie709815ab9bedb3d682e096b4005fdfb5e94d3a7
[ROCm/ROCR-Runtime commit: 5131ab4e64]
Try to reserve as much SVM address space as GPUVM can address.
Implement a fallback scheme to smaller sizes if larger allocations
fail or are not addressable by the GPU, down to an (arbitrary)
minimum of 4GB.
Change-Id: I770177834cc9e6ddd6ef4f20d789eab63c8055cb
[ROCm/ROCR-Runtime commit: 39bde26c9b]
Kernel ioctl AMDKFD_IOC_GET_PROCESS_APERTURES returns process apertures
only for GPU nodes. The current implementation assumed that this list of
GPU nodes returned by the ioctl has one to one correpondence to sysfs
topology nodes. This fails when non-GPU nodes exist in topology as in
case of Intel + gfx802
Fix this by using gpu_id (./sys/.../kfd/topology/nodes/1/gpu_id) to map
information obtained from kernel ioctl call.
Change-Id: I4ab8ae5354f12cf0b6609fc4b24182b82eb3677f
[ROCm/ROCR-Runtime commit: 5cc56a2647]
Fix TONGA_PAGE_SIZE value and move it to libhsakmt.h for usiing it
consistently in all places that require the same alignment for the
same reason. Create a generic alignment helper macro to replace some
incorrect hand-coded size alignments.
Move virtual address and size alignments down into aperture management
functions. Alignment is a per-aperture property that is set during
fmm_init_process_apertures. Doing the alignment there ensures that
all allocations in the same aperture are aligned the same way. Finding
objects by size and address can take the alignment into account.
Also align the size of physical allocations to back aligned virtual
address allocations. CPU mappings do not need to be aligned.
Map anonymous pages over released memory mappings to allow the
backing pages to be released, while keeping the address space
reserved.
Add alignment parameter to free_exec_aligned_memory_gpu to match the
interface of allocate_exec_aligned_memory_cpu. It doesn't make sense
to allow an alignment parameter in one but assume a specific
alignment in the other.
Change-Id: I74226ca6938f4948f643e5aee1d474720cd89e78
[ROCm/ROCR-Runtime commit: 6a5ca4bc5a]
The default is non-coherent access for better performance on dGPU.
Disabled hsaKmtSetMemoryPolicy function on dGPU to prevent app from
overriding the APE1 settings at runtime.
Fixed dGPU VM aperture limit to be inclusive.
Change-Id: I378ff74a654f533572775c0c97c19779a56bc6d9
[ROCm/ROCR-Runtime commit: 8e836f8183]
Currently, Kernel imposes a limit on VM. Thunk should be aware of it.
This fix is required till Kernel VM limit is sorted.
For now both "Host Access" memory and "Local Memory" share the same VM range.
Change-Id: I5a9220face20df9ede2b78bd6201a01dd2ea70e0
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 1438f15fd0]
Memory size is 64-bit. So use HSAuint64 instead of uint32.
Change-Id: Iaa607dec9c1a1c5ac46ea442fd482210ea550b45
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 4b768872c0]
Patch submitted by Besar Wicaksono
1. Bug on detecting local memory size interpreted as 32 bit value
instead of 64. The bug causes thunk to go into an infinite loop trying
to reserve virtual address range for dgpu system memory.
2. SIMD count in the node property is 0. Runtime use this attribute to
find a gpu device.
Regarding other attributes of intel+tonga topology, Harish started a
discussion on August iirc, could you please share an update ?
This would help me progress with more tests such as scratch memory,
which require the scratch aperture information in order to construct a
buffer srd in gpuvm space.
3. Bug on releasing memory via fmm_release, where no actual release is
being done. The vm_object can't be found because the memory size does
not match due to the allocation padded the size with 32KB.
4. Pointer arithmetic on vm_area allocation/release. The value of
vm_area_t::end seems to be interpreted inconsistently whether it is
(start + size -1) or (start + size).
One example of potential issue I see is the logic could report
larger size of the hole in the vm area list.
5. Resource cleanup on multiple library load/unload within a single
process.
- Any memory allocation on subsequent library load will result
an error "va above limit". To my understanding this is due to
the reserved memory for the system memory not being released on unload.
- The static variable events_page needs to be invalidated
appropriately on library unload so the next load could
reinitialize it.
6. Could you please update if AQL queue is ready to test with the stg
kfd/kmt ?
7. The system memory allocation with size larger than 32KB seems to be
padded by an extra 32KB. I was wondering if we could remove this
overhead.
Change-Id: I039988d36637525089c7569dc3b77e58750e2121
[ROCm/ROCR-Runtime commit: ee08f537a7]
__fmm_release expects gpu_id as parameter. fmm_release passed it instead the
index of the GPU in the gpu_mem array.
Signed-off-by: Oded Gabbay <oded.gabbay@amd.com>
[ROCm/ROCR-Runtime commit: aba396bc52]
Felix Kuehling
Ben Goz
shaoyunl
Amber Lin
Harish Kasiviswanathan
Yair Shachar
Moses Reuben
Oded Gabbay
Alexey Skidanov